Research On The Matadata Organization Method Of Multi-source And Heterogeneous Network-aware Information

With the increasing use of network awareness,there is a massive and multi-source heterogeneous network-aware information,the perceived data resources and decentralized data sources are also more and more,they will bring a strong impetus to network innovation and application innovation,How to efficiently organize and manage the acquirednetwork-aware information to improve the effective use has become the problems to be solved.For multi-source and heterogeneous data sets,data virtualization has achieved a unified data accessand integration based on metadata,therefore,It is also applicable to the integrated management of multi-source heterogeneous network-aware information.At present,Data virtualization system organizes the metadata using the traditional tree structure,it does not consider the relationships of metadata among different data sources.With the continuous increasing of network-aware information,the height of the metadata tree will continue to grow,and the query efficiency will be getting lower obviously.Based on data virtualization,this paper studies a metadata clustering organization method.The main work and contribution of this paper are listed below:Firstly,aiming to solve the problem of heterogeneity from multi-source network-aware information,a unified description method of network-aware information is studied and designed.The label format and structure of network-aware information is designed by the XML language,and the label <information></ information> represents a network-aware information,which includes the five parts of its category names,sources,storage locations,specific descriptions and related parameters.Secondly,the query efficiency of the tree structure used for the metadata organization in the current data virtualization system is low,considering that the metadata belonging to same categorycan be organized uniformly,a method of metadata clustering organization of network-aware information is proposed,which can effectively shorten the height of tree structure and improve the query efficiency.In the metadata clustering organization,each cluster corresponds to a network-aware information class.In order to effectively obtain the cluster and its number of the data sources,In this paper,the classical hierarchical clustering algorithm is used to cluster the XML documents of different types of network-aware informationfrom different sources.Then,the clustering results areuploaded to the data virtualization system whichcan automatically extract the metadata of the data source to realize the clustering organization of the metadata based on network-aware information class.Finally,for the mobile Internet-aware information obtained from the WAP gateway,a data virtualization prototype system is designed and built,and it could achieve the unified access and integration of multi-source and heterogeneous data.In addition,the performance test of the metadata clustering organization method of network-aware information proposed in this paper is carried out emphatically and compared with the traditional tree structure currently used in the data virtualization system.The results show that the method proposed in this paper has an average rate of precision increased by10.3% and also has a certain improvement in the efficiency of the query while ensures the recall ratio of the query.